Full-band super-resolution reconstruction method for hyperspectral image

Abstract

The invention discloses a full-band super-resolution reconstruction method for a hyperspectral image, and relates to the field of remote sensing image processing. Dividing wavebands by analyzing the spectral ranges of the obtained HS remote sensing image and MS remote sensing image; the wavebands with the same spectral range are defined as matched wavebands, and the wavebands with different spectral ranges are defined as non-matched wavebands; for the wave band matching part, a wavelet transform method is adopted for fusion processing, and the texture details and other aspects of the fused reconstructed image are clearer; for a non-matching part, a BP neural network is constructed by using the front and back changes of a pixel value between a reconstruction result of a matching wave band and an original HS remote sensing image as training knowledge. Pixels of a non-matched band HS remote sensing image are used as input of a network, a simulation result is optimized in combination with full-band normalized information entropy of the HS remote sensing image and a neighborhood spatial relationship of the pixels, full-band super-resolution reconstruction of the HS remote sensing image is realized, and the reconstructed HS remote sensing image has high spatial resolution of an MS remote sensing image and has high resolution. And the high spectral resolution is kept.

Description

technical field [0001] The invention relates to the field of remote sensing image processing, in particular to a hyperspectral image full-band super-resolution reconstruction method. Background technique [0002] Hyper-Spectral (HS) images collect hundreds of adjacent bands, providing fine spectral details of different objects. With its high spectral resolution, HS remote sensing images have good application prospects in many fields such as ground object recognition and ground target classification, environmental monitoring, land use and land cover change detection. However, the low spatial resolution of HS images severely limits its spatial fine-grained interpretation capability. Usually, improving its spatial resolution can be achieved from two aspects of remote sensing equipment and image processing. The former is always technically much more difficult and more expensive than the latter. [0003] Domestic and foreign researchers have done a lot of work in the field of ...

AI Technical Summary

Problems solved by technology

However, there is relatively little research on the reconstruction or simulation of HS remote sensing images outside the range of multispectral bands.

Images